speech
speech is a poor signal to noise
overcoming the poor signal-phonemic restoration effect
context allows brain to predict what phoneme should have been there and makes you actually hear it-fires that phoneme detector-only works when the noise would have masked the phoneme
how do we know it actually fires the phoneme detector?
run a study w ambiguous sound (dime/time ex.)
Adapt to “Tony the Tiger” or “Tony the *ger” or “Tony the _ger” (where * is cough / is blank).
If cough the missing phoneme adapts (Samuel, ’81)
aids help disambiguate phonemes
visual input
McGurk effect: people integrate what they see the mouth do with the sound that hits the ears
dubbing experiments
voice says “ba” and lips go “ga”, people hear “da”
• Lip movement for “ga” – similar (visual) movement to “da” not “ba”
• In terms of sound “ba” is more similar to “da” than “ga”.
• So, brain takes the “best” choice and you hear “da” – sound that is close to what is heard and matches lip movement -
easier to understand speaker when you can see them talk
gestures also help - congenitally blind people gesture
parsing phonemes is difficult
coarticulation
allows for super fast speaking - but means we blend language together
listening to spoken language is like reading text where the words overlap
ex. hearing unfamiliar language - seems like no break between words
if look at sonogram of our speech there are no breads between words - brain inserts breaks
how?
top down influences
knowledge of structural rules/statistical regularities (ex. words can end but not begin with “RK”
context influences how we perceive phonemes
ex. words in and out of sentence
much easier to identify words in sentences
ex. pollack and picket - tapes Ss in waiting room, then isolated single words and played them back
people could not identify 50% of their own words
two influences of context
semantic context allows brain to constrain appropriate words
speech context (how speaker talks) allows brain to undo effects of coarticulation
Top down influences on speech perception
Context & Expectation have large effects-
Our brains parse information into meaningful parts and insert breaks where there are none
phonetic restoration -
our brains reinsert information that is masked, so we still hear it
problems w understanding a sentence:assuming you can get the right words…how do you extract the right meaning?
ex. he made her duck
surface structure - the words on the page
deep structure - the meaning (semantics)
same surface structures can have multiple deep structures
context deals with ambiguity
speech is not a great signal, but we have developed many ways to constrain interpretations to help disambiguate